Oricom Technologies
www.oricomtech.com

Snail Vision

Return to:
Projects page.

This page was inspired by Alice.C and her narrations about her pet snail Speedy, with the snail equivalent of a 140 IQ. Alice claims the snail has learned to recognize her, and jumps out of its tank to show its emotion.

Original picture from The Apple Snail website, which says that snails have 44 pixel resolution in their visual systems. This probably means a 6x7 grid of photoreceptors, but this does not provide very useful vision, as shown by some of the pictures below. Therefore, we used mostly 40x40 pixel resolution here, as something that might be of practical interest for low-res robotic vision.


original - 200x165 pixels

grey scale - 256 levels

what they said 40 pixel res looks
like -(really just hi-res heavily blurred)

actual sample of above at 40x40
pixel resolution

The following images show the results of some simple image-processing transformations on the 40x40 low-res image above. These transforms are based upon the idea that individual visual cells in animals typically have "antagonistic center-surround" receptive fields, where light shown in areas surrounding a central region produce opposite effects to light shown in the central regions. For more info, see: The Computational Eye.


enhance detail

sharpen medium

sharpen heavy

edge enhance

edge detect light

edge detect medium

Alice's kid - what Alice thinks
her snail is seeing

what the snail is likely seeing
- Calvin [left image] resampled
at 6x6 res

original shell resampled at 6x6
res (what the snail likely sees)

For comparison purposes, the following images show the same transforms done on a blurred image. It would appear that heavy blurring may not be as detrimental for low-res vision as one might guess a priori. In other words, a robot using a low-res visual system may not need to worry too much about precisely focussing the images.

Another way of looking at this is that, the images above contain a lot of noise, and it may help successive processing (eg, area boundary definition, continuous surface identification, etc) if this noise were first removed (smoothed). This more or less can be seen by viewing the above images while squinting the eyes.


grey scale - 256 levels

blurred grey scale - 200x165 pixels

blurred - resampled at 40x40 pixels

enhance detail

sharpen medium

sharpen heavy

edge enhance

edge detect light

edge detect medium

<| TOP


© Oricom Technologies, Sep 2001